GitHub Copilot Coding Agent: Resolve Merge Conflicts via @copilot

Copilot Can Now Resolve Merge Conflicts on Pull Requests

GitHub has expanded the capabilities of the Copilot coding agent to include autonomous merge conflict resolution on pull requests. The feature allows developers to delegate one of the most tedious recurring tasks in collaborative development — untangling conflicting branches — directly to Copilot.

How It Works

To trigger the agent, a developer leaves a comment on any pull request mentioning @copilot with a natural-language instruction. A typical invocation looks like: "@copilot Merge in main and resolve the conflicts." From that point, the Copilot coding agent takes over in its cloud-based environment. It merges the target branch, resolves the conflicts, verifies that the repository's CI builds and tests still pass, and pushes the result back to the PR branch — all without the developer needing to pull down the branch locally.

The agent's conflict resolution operates within the same sandboxed cloud environment used for other agentic coding tasks, meaning it can run build scripts and test suites as part of its verification step before committing.

Availability and Plan Requirements

The merge conflict resolution capability is available to all users on paid Copilot plans: Copilot Pro, Pro+, Business, and Enterprise. It is not available on the free Copilot tier.

This feature builds directly on the @copilot-on-any-PR capability that GitHub shipped two days prior, which opened up the Copilot coding agent to work on pull requests beyond just those it originally created. Merge conflict resolution is one of the first concrete high-value use cases enabled by that foundation.

Why This Matters

Merge conflicts have historically required manual intervention: pulling the branch locally, running git merge, carefully editing conflicting sections, re-running tests, and pushing back up. For large codebases with active main branches, this can happen frequently and consume meaningful developer time.

By offloading this to Copilot, developers can stay in their current task flow and simply review the agent's resolution rather than performing it themselves. The cloud-based execution also means the developer's local environment is never interrupted.